Re: divison by zero trap
On Thu, 22 Jan 1998, Guenter Geiger wrote:
> Looking through the axp-kernel archives I found the solution for the
> problem. There's a system call for that. It is
>
> ieee_set_fp_control(ctrl_bits);
>
> the argument is a bit map see FPCR_*** in include/asm-alpha/fpu.h.
>
> With this call you can enable trapping on a per thread basis.
>
> Somebody suggested that the default behaviour should be as on Intel
> (trapping disabled for division by zero)
> so this call should probably be included into glibc. (probably this is
> done already and my glibc version is too old ?)
That's probably not a bad idea :) Michael Dorman just released a new
glibc for the Alpha. I'm not sure if it includes this or not, though
(probably not). I'll look since I just installed his new glibc into my
system last night.
> Well, I don't know enough to say something about this, this was just
> a suggestion, because more people will run into this problem,
> especially when porting math packages.
Yeah, that's a good idea. I found the same thing when I poked into the
kernel source last night. I'll also have to check to make sure that it's
the same throughout the development and stable kernels, btw.
C
--
TO UNSUBSCRIBE FROM THIS MAILING LIST: e-mail the word "unsubscribe" to
debian-alpha-request@lists.debian.org .
Trouble? e-mail to templin@bucknell.edu .
Reply to: